Reinforcement learning finds explicit graph realizations for three of six previously unresolved extreme rays of the N=6 holographic entropy cone and supplies evidence that the other three lie outside it.
Branes with Brains: Exploring String Vacua with Deep Reinforcement Learning
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
abstract
We propose deep reinforcement learning as a model-free method for exploring the landscape of string vacua. As a concrete application, we utilize an artificial intelligence agent known as an asynchronous advantage actor-critic to explore type IIA compactifications with intersecting D6-branes. As different string background configurations are explored by changing D6-brane configurations, the agent receives rewards and punishments related to string consistency conditions and proximity to Standard Model vacua. These are in turn utilized to update the agent's policy and value neural networks to improve its behavior. By reinforcement learning, the agent's performance in both tasks is significantly improved, and for some tasks it finds a factor of O(200) more solutions than a random walker. In one case, we demonstrate that the agent learns a human-derived strategy for finding consistent string models. In another case, where no human-derived strategy exists, the agent learns a genuinely new strategy that achieves the same goal twice as efficiently per unit time. Our results demonstrate that the agent learns to solve various string theory consistency conditions simultaneously, which are phrased in terms of non-linear, coupled Diophantine equations.
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
PPO reinforcement learning accelerates identification of gravitational wave signals from supercooled phase transitions in a minimal dark U(1)_x sector compared to Monte Carlo sampling.
In a wafer-modified Bousso-Polchinski model, 99.95% of the 532 million Calabi-Yau fourfold configurations in the Schöller-Skarke database allow vacuum energy spacings of 10^{-120} or smaller, with membrane nucleation transitions dominated by giant flux leaps under thin-wall approximations.
A Transformer trained by reinforcement learning generates heterotic line-bundle sums that satisfy anomaly-cancellation, stability, and chirality constraints, and its policy transfers usefully across Calabi-Yau geometries.
Neural networks classify Seiberg dual classes on Z_m x Z_n orbifolds with R^2=0.988 and predict toric multiplicities for Y^{6,0} with mean absolute error 0.021 under fixed Kasteleyn representative.
Machine learning optimization of a generalized SU(5) parameter y finds y ≈ 0.8 produces the closest match to the original model while resolving the fermion mass discrepancy.
The authors use Adam optimization over a 33-dimensional parameter space in an SU(5) model with 45 and 45bar Higgs representations to search for Yukawa values that make the proton lifetime exceed the Super-Kamiokande bound of 5.9e33 years.
citing papers explorer
-
Exploring the holographic entropy cone via reinforcement learning
Reinforcement learning finds explicit graph realizations for three of six previously unresolved extreme rays of the N=6 holographic entropy cone and supplies evidence that the other three lie outside it.
-
Supercool with PPO: Exploring Supercooled Phase Transitions via Reinforcement Learning
PPO reinforcement learning accelerates identification of gravitational wave signals from supercooled phase transitions in a minimal dark U(1)_x sector compared to Monte Carlo sampling.
-
Small Vacuum Energy and Tunneling in a Modified Bousso-Polchinski Model
In a wafer-modified Bousso-Polchinski model, 99.95% of the 532 million Calabi-Yau fourfold configurations in the Schöller-Skarke database allow vacuum energy spacings of 10^{-120} or smaller, with membrane nucleation transitions dominated by giant flux leaps under thin-wall approximations.
-
Exploring Line Bundle Standard Models with Transformers
A Transformer trained by reinforcement learning generates heterotic line-bundle sums that satisfy anomaly-cancellation, stability, and chirality constraints, and its policy transfers usefully across Calabi-Yau geometries.
-
Machine Learning Toric Duality in Brane Tilings
Neural networks classify Seiberg dual classes on Z_m x Z_n orbifolds with R^2=0.988 and predict toric multiplicities for Y^{6,0} with mean absolute error 0.021 under fixed Kasteleyn representative.
-
Good flavor search in SU(5): a machine learning approach
Machine learning optimization of a generalized SU(5) parameter y finds y ≈ 0.8 produces the closest match to the original model while resolving the fermion mass discrepancy.
-
Optimizing Yukawa couplings to suppress Dimension-five Proton Decay in $SU(5)$ GUT
The authors use Adam optimization over a 33-dimensional parameter space in an SU(5) model with 45 and 45bar Higgs representations to search for Yukawa values that make the proton lifetime exceed the Super-Kamiokande bound of 5.9e33 years.